OBJECTIVE

To assess the association between use of sodium–glucose cotransporter 2 (SGLT2) inhibitors and the risk of new-onset atrial fibrillation (AF) in routine clinical practice.

RESEARCH DESIGN AND METHODS

We used nationwide registers in Denmark, Norway, and Sweden from 2013 to 2018 in order to include patients without a history of AF who were newly prescribed an SGLT2 inhibitor or an active comparator (glucagon-like peptide 1 [GLP-1] receptor agonist). We performed a cohort study to assess new-onset AF in intention-to-treat analyses using Cox regression, adjusted for baseline covariates with propensity score weighting.

RESULTS

We identified 79,343 new users of SGLT2 inhibitors (59.2% dapagliflozin, 40.0% empagliflozin, 0.8% canagliflozin, <0.1% ertugliflozin) and 57,613 new users of GLP-1 receptor agonists. Mean age of the study cohort was 61 years and 60% were men. The adjusted incidence rate of new-onset AF was 8.6 per 1,000 person-years for new users of SGLT2 inhibitors compared with 10.0 per 1,000 person-years for new users of GLP-1 receptor agonists. The adjusted hazard ratio (aHR) was 0.89 (95% CI 0.81–0.96), and the rate difference was 1.4 fewer events per 1,000 person-years (95% CI 0.6–2.1). Using an as-treated exposure definition, the aHR for new-onset AF was 0.87 (95% CI 0.76–0.99). No statistically significant heterogeneity of the aHRs was observed between subgroups of patients with and without a history of heart failure or major cardiovascular disease.

CONCLUSIONS

In this cohort study using nationwide data from three countries, use of SGLT2 inhibitors, compared with GLP-1 receptor agonists, was associated with a modestly reduced risk of new-onset AF.

Atrial fibrillation (AF) is the most prevalent sustained arrhythmia with a large impact on morbidity and mortality (1,2). Patients with type 2 diabetes experience an increased risk of AF (35), which is possibly mediated through type 2 diabetes–related comorbidities, including obesity, hypertension, chronic kidney disease, and heart failure. Moreover, type 2 diabetes impacts myocardial remodeling and electrical properties of the heart, potentially increasing the propensity to develop AF (6,7).

Sodium–glucose cotransporter 2 (SGLT2) inhibitors are a class of glucose-lowering drugs that act by increasing urinary glucose excretion (8). Large randomized controlled trials have shown that SGLT2 inhibitors have beneficial effects on several conditions associated with AF, including atherosclerotic cardiovascular events, heart failure–related outcomes, progression of chronic kidney disease (912), and reduction of body weight and blood pressure (13,14). In addition, experimental studies indicate that SGLT2 inhibitors might protect against cardiac fibrosis, electrical remodeling, and myocardial inflammation, and also lower urate levels and improve hemodynamics (1518). It has therefore been suggested that SGLT2 inhibitors could protect against AF (19,20).

In a meta-analysis of data from randomized controlled trials (19), including the Dapagliflozin Effect on Cardiovascular Events trial (DECLARE-TIMI 58) (20), patients who received an SGLT2 inhibitor had a lower risk of AF events compared with those who received placebo or other glucose-lowering drugs. However, the interpretation of these findings is uncertain as they were based on post hoc analyses of AF events that were not systematically collected, had varying definitions, and depended on active reporting by investigators. Furthermore, the clinical trials were performed in patients with established cardiovascular disease or a high cardiovascular risk, and it is not clear whether the findings can be generalized to the broader group of patients seen in routine clinical practice (21). Previous observational studies assessing the association between use of SGLT2 inhibitors and AF have suffered from substantial limitations, including immortal time bias and time-lag bias (22,23). We performed a register-based cohort study of nationwide data in Sweden, Denmark, and Norway to investigate the risk of new-onset AF among new users of SGLT2 inhibitors versus an active comparator, glucagon-like peptide 1 (GLP-1) receptor agonists, in routine clinical practice.

Study Design

We used data from nationwide health and administrative registers in Sweden, Denmark, and Norway from April 2013 through December 2018. The registers included population registers (vital status, demographics), Statistics Denmark/Sweden (socioeconomic variables), patient registers (comorbidities, outcomes), prescription registers (study drugs, comedications), the Swedish National Diabetes Register (glycated hemoglobin level, blood pressure, albuminuria, estimated glomerular filtration rate [eGFR], BMI, smoking), and the Danish Register of Laboratory Results for Research (glycated hemoglobin, albuminuria, eGFR) (details provided in the Supplementary Material).

We used an active comparator, new user design to mitigate the risk for confounding by indication and unmeasured confounding (24). We used GLP-1 receptor agonists as the active comparator because clinical guidelines used during the study period recommended SGLT2 inhibitors and GLP-1 receptor agonists as second- or third-line glucose-lowering therapies (with both drug classes likely being considered for patients at high cardiovascular risk), the two drug classes show similar temporal trends of use, and available data indicate that GLP-1 receptor agonists have a neutral effect on study outcomes (25,26).

Study Population

All patients in the three countries aged 35–84 years who were new users of either an SGLT2 inhibitor or a GLP-1 receptor agonist were eligible for inclusion. The upper age limit was used to limit the potential impact of frailty, and the number of patients initiating SGLT2 inhibitors and GLP-1 receptor agonists at age >85 years was also very limited in our data. The anatomic therapeutic chemical codes for the study drugs are provided in Supplementary Table 1. New use was defined as no use of either study drug class at any time before cohort entry. The date of filling the first prescription constituted cohort entry.

While type 2 diabetes was not an explicit cohort eligibility criterion due to diagnosis in primary care not being covered by the national patient registries, during the study period, SGLT2 inhibitors were not indicated for heart failure without concomitant diabetes, and the large randomized controlled trials providing evidence for a beneficial effect on heart failure regardless of diabetes status had not been published. Therefore, it can be assumed that new users of SGLT2 inhibitors included in the study were prescribed the drug on the indication of type 2 diabetes.

We excluded patients with a history of AF at any time before cohort entry or patients who used oral anticoagulants within the past year. We further excluded patients who had a history of dialysis or renal transplantation, end-stage illness, drug misuse, severe pancreatic disorders, hospital admission for any reason within 30 days before cohort entry, no specialist care contact or use of any prescription drug in the previous year, or use of liraglutide with obesity indication (Supplementary Table 2).

In the pooled data set of patients from the three countries, logistic regression was used to estimate a propensity score representing the probability of starting an SGLT2 inhibitor versus a GLP-1 receptor agonist, conditional on the status of 61 covariates at cohort entry (Supplementary Table 3). The variables included sociodemographic characteristics, comorbidities, comedications, health care utilization (including hospitalizations and specialist outpatient care visits) and two-way interaction terms between country and each covariate. Missing data on place of birth (0.1%) and civil status (<0.5%) were handled with use of missing categories (27).

We used inverse probability of treatment weighting based on the propensity score (average treatment effect weighting) to control for confounding (28). For subgroup analyses (described below), a separate propensity score was estimated within each subgroup. Patients with a propensity score outside the overlapping area of the propensity score distributions were excluded. We used stabilized weight to mitigate the impact of extreme weights.

We used an intention-to-treat exposure definition in which patients remained on their initial exposure status throughout follow-up. Patients were followed from treatment initiation to outcome event, emigration, death, 5 years of follow-up, or end of the study period.

Outcome

The outcome was a first diagnosis of AF or atrial flutter (ICD-10 code I48) captured in the patient registers during any type of health care contact (planned or emergency outpatient visit or hospitalization, primary or secondary diagnosis). Scandinavian validation studies have shown that the positive predictive value for the diagnosis of AF in the national registers is >95% (29,30).

Statistical Analyses

Cox proportional hazards regression with time since start of treatment as the time scale was used to estimate hazard ratios (HRs) for use of SGLT2 inhibitors versus GLP-1 receptor agonists. The 95% CIs that did not overlap with 1 were considered statistically significant. Covariate balance between the study groups was assessed using standardized differences, and values <0.10 were considered indicative of good balance. We described the cumulative incidence using Kaplan-Meier curves. All results presented as adjusted are from the propensity score–weighted analyses.

We conducted prespecified subgroup analyses by age-group (35–64 years and ≥65 to 84 years), sex, history of major cardiovascular disease, and history of heart failure (Supplementary Table 4). Effect modification by subgroup status was examined with an interaction term between treatment status and subgroup; in these analyses, P < 0.05 was considered as statistically significant. We also conducted analyses by country to assess consistency across data sources and separate analyses for the individual SGLT2 inhibitors empagliflozin and dapagliflozin; this analysis was not performed for canagliflozin and ertugliflozin, as few patients used those drugs.

We performed prespecified additional analyses. First, we used an as-treated exposure definition wherein patients were considered exposed to the study drug as long as the prescriptions were refilled within the estimated duration of the most recent prescription plus a 30-day grace period, which was used to account for irregular drug use. Patients were censored at treatment cessation or crossover to the other study drug. Next, we performed an analysis in which follow-up time was divided into distinct categories in relation to cohort entry (<1 year, 1 to <3 years, ≥3 years).

We conducted prespecified sensitivity analyses. First, in the Swedish and Danish parts of the cohort, we used a propensity score, with additional variables providing information about disease severity and comorbidities, including glycated hemoglobin level, blood pressure, albuminuria, eGFR, BMI, and smoking in Sweden, and glycated hemoglobin level, albuminuria, and eGFR in Denmark (Supplementary Table 5). Given the proportion of missing values for the additional variables (Supplementary Table 5), multiple imputation (fully conditional specification imputation) with 10 imputed data sets was used (31). Imputation was based on all variables included in the propensity score, the additional variables, and the outcome variable. Second, in the Swedish and Danish parts of the cohort, we also included education in the propensity score. Third, since inverse probability of treatment weighting could generate large weights for patients with low probability of study drug exposure, we performed analyses with truncated weights (values >5 were set to 5), and analyses in which patients with weights in the highest and lowest 2.5 percentiles were excluded. Finally, the E-value was estimated to address the effect of potential unmeasured confounding. The E-value is defined as the minimal strength of association that an unmeasured confounder would have with the exposure and the outcome to explain away the observed association (32).

To investigate the association between use of SGLT2 inhibitors and AF in the overall population of patients initiating these drugs, we constructed an additional cohort using the same inclusion and exclusion criteria as in the primary analysis (Supplementary Table 2) with the exception that patients with a history of AF at baseline or who had used oral anticoagulants within the past year were not excluded. Hence, this broader study population included patients with and without a history of AF. In this analysis, the outcome was hospitalization with a primary diagnosis of AF. We considered this cohort to represent the routine clinical practice equivalent to the populations investigated in trials that reported AF outcomes and the outcome as an AF analog to the frequently used outcome of hospital admission for heart failure. In this cohort, we also conducted the subgroup analyses (with the addition of an analysis by history of AF at cohort entry) as well as additional and sensitivity analyses as described above.

The study was approved by the regional ethics committee in Stockholm, Sweden, and the regional committee for medical and health research ethics in Norway. In Denmark, ethics approval is not required for register-based research.

Study Population

We identified 79,343 new users of SGLT2 inhibitors and 57,613 new users of GLP-1 receptor agonists who were eligible for the analyses (Fig. 1). Population characteristics before and after propensity score weighting are shown in Table 1. Covariates in the two groups were well balanced after weighting. The median follow-up time was 1.6 (interquartile range [IQR] 0.7–2.8) years for users of SGLT2 inhibitors and 2.3 (IQR 1.1–3.8) years for users of GLP-1 receptor agonists. The proportion of total follow-up time by type of SGLT2 inhibitor was 59.2% for dapagliflozin, 40.0% for empagliflozin, 0.8% for canagliflozin, and <0.1% for ertugliflozin.

Primary Analysis

Figure 2 shows the adjusted cumulative incidence of new-onset AF. During follow-up, 1,273 users of SGLT2 inhibitors and 1,361 users of GLP-1 receptor agonists received a first diagnosis of AF. The adjusted incidence rate for new-onset AF was 8.6 per 1,000 person-years with use of SGLT2 inhibitors and 10.0 per 1,000 person-years with use of GLP-1 receptor agonists. The adjusted HR (aHR) was 0.89 (95% CI 0.81–0.96), and the rate difference was −1.4 events per 1,000 person-years (95% CI −2.1 to −0.6).

Subgroup and Additional Analyses

The results of the subgroup and additional analyses are shown in Fig. 3. The aHRs were not significantly different in subgroups by age. In the subgroup analysis by sex, the aHR for new-onset AF among women was 0.99 (95% CI 0.85–1.15) compared with 0.86 (95% CI 0.78–0.95) among men, although there was no statistical evidence of heterogeneity (P for interaction = 0.14). Incidence rates of new-onset AF were markedly higher among patients with a history of heart failure or major cardiovascular disease compared with those without such a history, but no statistically significant heterogeneity of the aHRs was observed. The aHRs were also similar for patients initiating empagliflozin and those initiating dapagliflozin. The results for analyses in each country are shown separately in Supplementary Table 6.

In the additional analysis using an as-treated exposure definition, the median follow-up time was 0.6 (IQR 0.3–1.1) years for users of SGLT2 inhibitors and 0.6 (IQR 0.2–1.2) years for users of GLP-1 receptor agonists. The aHR for new-onset AF, comparing SGLT2 inhibitors with GLP-1 receptor agonists, was 0.87 (95% CI 0.76–0.99). In the additional analysis stratified by time since treatment initiation (<1 year, 1 to <3 years, and ≥3 years), aHRs were consistent with that estimated using data from the entire duration of follow-up.

Sensitivity Analyses

In the analyses adjusted for additional variables (glycated hemoglobin, blood pressure, albuminuria, eGFR, BMI, and smoking) in the Swedish part of the cohort (patient characteristics shown in Supplementary Table 7), the aHR (0.90 [95% CI 0.79–1.01]) was similar to that of the country-specific analysis without such adjustments (0.85 [95% CI 0.75–0.97]) (Supplementary Table 8). Similarly, in the Danish part of the cohort (patient characteristics shown in Supplementary Table 9), adjustment for additional variables (glycated hemoglobin, albuminuria, and eGFR) resulted in an aHR (0.92 [95% CI 0.79–1.08]) identical to that in the country-specific analysis without additional adjustment (0.92 [95% CI 0.78–1.09]) (Supplementary Table 8). The findings were consistent with that of the primary analysis when using a propensity score that included education level in the Swedish and Danish parts of the cohort, truncating weights >5 and excluding patients with weights in the highest and lowest 2.5 percentiles (Supplementary Table 10). The estimated E-value was 1.5 for the point estimate of the primary analysis and 1.25 for the upper CI limit; hence, to move the CI to include 1 would require unmeasured confounding to have a 1.25-fold association with both the outcome and the exposure.

Analyses of the Additional Overall Cohort of Patients With and Without AF

In the additional overall cohort that included patients with and without AF, 87,525 new users of SGLT2 inhibitors and 63,921 new users of GLP-1 receptor agonists were included (Supplementary Fig. 1). Population characteristics before and after weighting are shown in Supplementary Table 11. The adjusted incidence rate for AF hospitalization was 5.1 with use of SGLT2 inhibitors and 5.5 with use of GLP-1 receptor agonists. The aHR was 0.91 (95% CI 0.82–1.01), and the rate difference was −0.4 events per 1,000 person-years (95% CI −0.9 to 0.2). Subgroup and additional analyses are shown in Supplementary Fig. 2. Among the patients with a history of AF at baseline, the aHR was 1.01 (95% CI 0.87–1.18). In the additional as-treated analysis, the aHR was 0.82 (95% CI 0.70–0.96). The results for analyses in each country are shown separately in Supplementary Table 12. The findings were consistent across sensitivity analyses (Supplementary Tables 1316).

In this cohort study using nationwide registers from three countries to include patients seen in routine clinical practice, use of SGLT2 inhibitors, compared with GLP-1 receptor agonists, was associated with a 11% relative risk reduction of new-onset AF, corresponding to an absolute risk reduction of 1.4 events per 1,000 person-years. The association with a lower relative risk reduction of new-onset AF was consistent in subgroups of patients with and without major cardiovascular disease and with and without heart failure, although the absolute risk differences were of larger magnitude in the subgroup of patients with a history of heart failure.

Post hoc analyses and meta-analyses of randomized controlled trials have suggested that SGLT2 inhibitors may reduce the risk of AF, although data on AF events were not systematically collected and depended on active reporting by investigators. A meta-analysis of 31 randomized controlled trials reported a 25% relative risk reduction of serious AF events in patients who received an SGLT2 inhibitor compared with those receiving placebo or other glucose-lowering drugs. In the DECLARE-TIMI 58, which carried the most weight in the meta-analysis, randomization to dapagliflozin was associated with a significantly reduced risk of a reported AF event (HR 0.81 [95% CI 0.68–0.95]) (20). The effect was consistent in patients with and without a history of AF at baseline. Similarly, in the Canagliflozin and Renal Events in Diabetes With Established Nephropathy Clinical Evaluation (CREDENCE) trial, which enrolled patients with type 2 diabetes and chronic kidney disease, the point estimate indicated a protective effect of randomization to canagliflozin (HR 0.76 [95% CI 0.53–1.10]), although this association was statistically significant only among patients without a history of AF at baseline (HR 0.64 [95% CI 0.43–0.96]) (33).

While the large clinical trials included patients at high cardiovascular or renal risk, two observational studies have assessed the relationship between use of SGLT2 inhibitors and the risk of new-onset AF among patients seen in routine clinical practice. In a Taiwanese study including 15,606 patients using SGLT2 inhibitors, there was a strong association between use of SGLT2 inhibitors versus dipeptidyl peptidase 4 inhibitors and new-onset AF events (HR 0.61 [95% CI 0.50–0.73]) (22). In another Taiwanese study comparing ∼80,000 patients using SGLT2 inhibitors with patients using other glucose-lowering drugs, the HR for new-onset AF was 0.84 (95% CI 0.66–1.07) (23). However, both these studies used designs that introduced immortal time, which has a strong potential for bias toward protective associations in favor of SGLT2 inhibitors (34,35). Moreover, the use of any other glucose-lowering drug as the comparator may introduce time-lag bias, since many of the drugs are not used at similar stages of the disease as SGLT2 inhibitors (36).

Our study substantially expands on previous observational data and post hoc analyses of clinical trials regarding the risk of new-onset AF associated with use of SGLT2 inhibitors. We reduced the risk for confounding by restricting the analysis to new users of the study drugs and by using a propensity score model that included a broad range of patient characteristics. The use of nationwide registers from three countries to include ∼140,000 patients seen in routine clinical practice strengthens the generalizability of our findings. Moreover, estimates were consistent across several sensitivity analyses. We used GLP-1 receptor agonists as the active comparator drug class to reduce the risk of confounding by indication and unmeasured patient characteristics. During the study period, large cardiovascular outcome trials showing protective effects for both SGLT2 inhibitors and GLP-1 receptor agonists were presented, and clinical guidelines recommended use of these two drug classes in similar clinical situations and at a similar stage of type 2 diabetes (9,10,12,37). While the slight increase in heart rate observed with use of GLP-1 receptor agonists have raised concerns regarding an increased risk of tachyarrhythmias (38), several meta-analyses of randomized clinical trials have provided data against this concern, with current evidence pointing toward a neutral effect of GLP-1 receptor agonists on the risk of AF (25,26). However, even if GLP-1 receptor agonists were not risk neutral, the analysis would still reflect the head-to-head comparative effectiveness of SGLT2 inhibitors versus GLP-1 receptor agonists.

In the main analyses, we used an intention-to-treat exposure definition. Hence, we explicitly aimed to include all available person-time among patients initiating SGLT2 inhibitors, regardless of whether patients fully adhered or switched to an alternative treatment. The main analyses thus investigated the overall clinical effectiveness of initiating an SGLT2 inhibitor. In the additional analysis with an as-treated exposure definition, the results were largely similar (aHR 0.87 [95% CI 0.76–0.99] vs. 0.89 [95% CI 0.81–0.96] in the primary analysis).

In addition to the primary analyses of new-onset AF among patients without a history of AF, we performed an analysis that included both patients with and without a history of AF. In these analyses, a statistically significant association between SGLT2 inhibitor treatment and hospitalization for AF was not observed in the intention-to-treat analysis. However, the estimate tended toward a protective association (aHR 0.91 [95% CI 0.82–1.01]), and the association was statistically significant in the as-treated analysis (aHR 0.82 [95% CI 0.70–0.96]). Nonetheless, the point estimate indicated that use of SGLT2 inhibitors was not associated with a reduced risk of hospitalization for AF (aHR 1.01 [95% CI 0.87–1.18]) in the subgroup of patients with a history of AF at baseline, although there was no evidence of a statistically significant heterogeneity in the aHR compared with those without a history of AF. It can be hypothesized that potential mechanisms by which SGLT2 inhibitors may protect against incident AF may not reduce AF severity among those with prevalent AF, although post hoc analyses of clinical trials showed reduced HRs that were of similar magnitudes in patients with and without prevalent AF.

Our study has limitations. First, although ICD codes for cardiovascular outcomes in the Scandinavian patient registers have high validity (3941) and Swedish and Danish validation studies of the codes used for the outcome definition have shown positive predictive values of >95%, misclassification may have introduced bias. However, it is unlikely that such an outcome misclassification would differ between users of SGLT2 inhibitors and GLP-1 receptor agonists. Second, the exclusion criteria used to create a study population free of AF at cohort entry included a previous diagnosis of AF and use of oral anticoagulants in the previous year. The latter criterion is likely to have excluded some patients who used anticoagulants for conditions other than AF; this may marginally limit the generalizability of the results. Third, the exposure definition was based on filled prescriptions; thus, low adherence may bias the results toward the null. Fourth, patients who initiated dapagliflozin or empagliflozin contributed most of the follow-up time, and the results mainly apply to these specific SGLT2 inhibitors. Fifth, median follow-up time among users of SGLT2 inhibitors in this study was 1.6 years, with 25% of the patients followed for ≥2.8 years. While these data reflect real-world utilization patterns during the study period, the accumulation of additional routine care data will allow future long-term evaluation of the association between SGLT2 inhibitor use and AF. Finally, although we used an active comparator, new user design and accounted for a wide range of patient characteristics, the observational nature of this study means that the possibility of residual and unmeasured confounding cannot be ruled out.In conclusion, in this cohort study using nationwide data from three countries, use of SGLT2 inhibitors, compared with GLP-1 receptor-agonists, was associated with a modestly reduced risk of new-onset AF.

This article contains supplementary material online at https://doi.org/10.2337/figshare.21562863.

Funding. The study was supported by grants from the Swedish Research Council, the Swedish Heart-Lung Foundation, the Swedish Diabetes Foundation, and Dr. Margaretha Nilsson’s Foundation for Medical Research. B.P. was supported by an investigator grant from the Strategic Research Area Epidemiology Program at Karolinska Institutet. P.U. was supported by a Karolinska Institutet faculty-funded career position. A.H. was supported by a Novo Nordisk Foundation investigator grant.

The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Duality of Interest. C.J. is an employee of NordicRWE. B.E. reports personal fees from Amgen, AstraZeneca, Boehringer Ingelheim, Eli Lilly, Merck Sharp & Dohme, Mundipharma, Navamedic, Novo Nordisk, and RLS Global and grants from Sanofi outside the submitted work. S.G. reports lecture fees and research grants from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Merck Sharp & Dohme, Novo Nordisk, and Sanofi outside the submitted work. H.S. reports consulting fees from Celgene and employment at IQVIA outside the submitted work. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. A.E., V.W., M.M., A.H., B.E., S.G., K.H., C.J., H.S., B.P., and P.U. contributed to the acquisition, analysis, or interpretation of data and critical revision of the manuscript for important intellectual content. A.E., V.W., B.P., and P.U. contributed to the study concept and design. A.E., B.P., and P.U. drafted the manuscript. V.W. conducted the statistical analysis. B.P. and P.U. obtained funding and provided study supervision. B.P. and P.U. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 13th Annual Meeting of the Nordic Pharmacoepidemiological Network, Stockholm, Sweden, 11–12 November 2021 and at the 38th International Conference for PharmacoEpidemiology, Copenhagen, Denmark, 24–28 August 2022.

1.
Rahman
F
,
Kwan
GF
,
Benjamin
EJ
.
Global epidemiology of atrial fibrillation
.
Nat Rev Cardiol
2014
;
11
:
639
654
2.
Odutayo
A
,
Wong
CX
,
Hsiao
AJ
,
Hopewell
S
,
Altman
DG
,
Emdin
CA
.
Atrial fibrillation and risks of cardiovascular disease, renal disease, and death: systematic review and meta-analysis
.
BMJ
2016
;
354
:
i4482
3.
Kannel
WB
,
Wolf
PA
,
Benjamin
EJ
,
Levy
D
.
Prevalence, incidence, prognosis, and predisposing conditions for atrial fibrillation: population-based estimates
.
Am J Cardiol
1998
;
82
(
7A
):
2N
9N
4.
Huxley
RR
,
Filion
KB
,
Konety
S
,
Alonso
A
.
Meta-analysis of cohort and case-control studies of type 2 diabetes mellitus and risk of atrial fibrillation
.
Am J Cardiol
2011
;
108
:
56
62
5.
Nichols
GA
,
Reinier
K
,
Chugh
SS
.
Independent contribution of diabetes to increased prevalence and incidence of atrial fibrillation
.
Diabetes Care
2009
;
32
:
1851
1856
6.
Russo
I
,
Frangogiannis
NG
.
Diabetes-associated cardiac fibrosis: cellular effectors, molecular mechanisms and therapeutic opportunities
.
J Mol Cell Cardiol
2016
;
90
:
84
93
7.
Watanabe
M
,
Yokoshiki
H
,
Mitsuyama
H
,
Mizukami
K
,
Ono
T
,
Tsutsui
H
.
Conduction and refractory disorders in the diabetic atrium
.
Am J Physiol Heart Circ Physiol
2012
;
303
:
H86
H95
8.
Kalra
S
.
Sodium glucose co-transporter-2 (SGLT2) inhibitors: a review of their basic and clinical pharmacology
.
Diabetes Ther
2014
;
5
:
355
366
9.
Zinman
B
,
Wanner
C
,
Lachin
JM
, et al.;
EMPA-REG OUTCOME Investigators
.
Empagliflozin, cardiovascular outcomes, and mortality in type 2 diabetes
.
N Engl J Med
2015
;
373
:
2117
2128
10.
Wiviott
SD
,
Raz
I
,
Bonaca
MP
, et al.;
DECLARE–TIMI 58 Investigators
.
Dapagliflozin and cardiovascular outcomes in type 2 diabetes
.
N Engl J Med
2019
;
380
:
347
357
11.
Perkovic
V
,
Jardine
MJ
,
Neal
B
, et al.;
CREDENCE Trial Investigators
.
Canagliflozin and renal outcomes in type 2 diabetes and nephropathy
.
N Engl J Med
2019
;
380
:
2295
2306
12.
Neal
B
,
Perkovic
V
,
Mahaffey
KW
, et al.;
CANVAS Program Collaborative Group
.
Canagliflozin and cardiovascular and renal events in type 2 diabetes
.
N Engl J Med
2017
;
377
:
644
657
13.
Cheong
AJY
,
Teo
YN
,
Teo
YH
, et al
.
SGLT inhibitors on weight and body mass: a meta-analysis of 116 randomized-controlled trials
.
Obesity (Silver Spring)
2022
;
30
:
117
128
14.
Mazidi
M
,
Rezaie
P
,
Gao
HK
,
Kengne
AP
.
Effect of sodium-glucose cotransport-2 inhibitors on blood pressure in people with type 2 diabetes mellitus: a systematic review and meta-analysis of 43 randomized control trials with 22 528 patients
.
J Am Heart Assoc
2017
;
6
:
e004007
15.
Nishinarita
R
,
Niwano
S
,
Niwano
H
, et al
.
Canagliflozin suppresses atrial remodeling in a canine atrial fibrillation model
.
J Am Heart Assoc
2021
;
10
:
e017483
16.
Kim
SR
,
Lee
SG
,
Kim
SH
, et al
.
SGLT2 inhibition modulates NLRP3 inflammasome activity via ketones and insulin in diabetes with cardiovascular disease
.
Nat Commun
2020
;
11
:
2127
17.
Bailey
CJ
.
Uric acid and the cardio-renal effects of SGLT2 inhibitors
.
Diabetes Obes Metab
2019
;
21
:
1291
1298
18.
Sano
M
.
Hemodynamic effects of sodium-glucose cotransporter 2 inhibitors
.
J Clin Med Res
2017
;
9
:
457
460
19.
Pandey
AK
,
Okaj
I
,
Kaur
H
, et al
.
Sodium-glucose co-transporter inhibitors and atrial fibrillation: a systematic review and meta-analysis of randomized controlled trials
.
J Am Heart Assoc
2021
;
10
:
e022222
20.
Zelniker
TA
,
Bonaca
MP
,
Furtado
RHM
, et al
.
Effect of dapagliflozin on atrial fibrillation in patients with type 2 diabetes mellitus: insights from the DECLARE-TIMI 58 trial
.
Circulation
2020
;
141
:
1227
1234
21.
Birkeland
KI
,
Bodegard
J
,
Norhammar
A
, et al
.
How representative of a general type 2 diabetes population are patients included in cardiovascular outcome trials with SGLT2 inhibitors? A large European observational study
.
Diabetes Obes Metab
2019
;
21
:
968
974
22.
Ling
AWC
,
Chan
CC
,
Chen
SW
, et al
.
The risk of new-onset atrial fibrillation in patients with type 2 diabetes mellitus treated with sodium glucose cotransporter 2 inhibitors versus dipeptidyl peptidase-4 inhibitors
.
Cardiovasc Diabetol
2020
;
19
:
188
23.
Chen
HY
,
Huang
JY
,
Siao
WZ
,
Jong
GP
.
The association between SGLT2 inhibitors and new-onset arrhythmias: a nationwide population-based longitudinal cohort study
.
Cardiovasc Diabetol
2020
;
19
:
73
24.
Lund
JL
,
Richardson
DB
,
Stürmer
T
.
The active comparator, new user study design in pharmacoepidemiology: historical foundations and contemporary application
.
Curr Epidemiol Rep
2015
;
2
:
221
228
25.
Monami
M
,
Nreu
B
,
Scatena
A
, et al
.
Glucagon-like peptide-1 receptor agonists and atrial fibrillation: a systematic review and meta-analysis of randomised controlled trials
.
J Endocrinol Invest
2017
;
40
:
1251
1258
26.
Hamedi
Z
,
Mishriky
BM
,
Okunrintemi
V
,
Powell
JR
,
Cummings
DM
.
GLP-1 RA and atrial fibrillation in the cardiovascular outcome trials
.
Diabetes Metab Res Rev
2021
;
37
:
e3436
27.
D’Agostino
RB
,
Rubin
DB
.
Estimating and using propensity scores with partially missing data
.
J Am Stat Assoc
2000
;
95
:
749
759
28.
Austin
PC
.
An introduction to propensity score methods for reducing the effects of confounding in observational studies
.
Multivariate Behav Res
2011
;
46
:
399
424
29.
Smith
JG
,
Platonov
PG
,
Hedblad
B
,
Engström
G
,
Melander
O
.
Atrial fibrillation in the Malmö Diet and Cancer study: a study of occurrence, risk factors and diagnostic validity
.
Eur J Epidemiol
2010
;
25
:
95
102
30.
Sundbøll
J
,
Adelborg
K
,
Munch
T
, et al
.
Positive predictive value of cardiovascular diagnoses in the Danish National Patient Registry: a validation study
.
BMJ Open
2016
;
6
:
e012832
31.
Harel
O
,
Zhou
XH
.
Multiple imputation: review of theory, implementation and software
.
Stat Med
2007
;
26
:
3057
3077
32.
VanderWeele
TJ
,
Ding
P
.
Sensitivity analysis in observational research: introducing the E-value
.
Ann Intern Med
2017
;
167
:
268
274
33.
Zhou
Z
,
Jardine
MJ
,
Li
Q
, et al.;
CREDENCE Trial Investigators
.
Effect of SGLT2 inhibitors on stroke and atrial fibrillation in diabetic kidney disease: results from the CREDENCE trial and meta-analysis
.
Stroke
2021
;
52
:
1545
1556
34.
Suissa
S
.
Lower risk of death with SGLT2 inhibitors in observational studies: real or bias?
Diabetes Care
2018
;
41
:
6
10
35.
Suissa
S
.
Reduced mortality with sodium-glucose cotransporter-2 inhibitors in observational studies: avoiding immortal time bias
.
Circulation
2018
;
137
:
1432
1434
36.
American Diabetes Association
.
9. Pharmacologic approaches to glycemic treatment: Standards of Medical Care in Diabetes-2021
.
Diabetes Care
2021
;
44
(
Suppl. 1
):
S111
S124
37.
Marso
SP
,
Daniels
GH
,
Brown-Frandsen
K
, et al
.
Liraglutide and cardiovascular outcomes in type 2 diabetes
.
N Engl J Med
2016
;
375
:
311
322
38.
Smits
MM
,
Tonneijck
L
,
Muskiet
MHA
, et al
.
Heart rate acceleration with GLP-1 receptor agonists in type 2 diabetes patients: an acute and 12-week randomised, double-blind, placebo-controlled trial
.
Eur J Endocrinol
2017
;
176
:
77
86
39.
Ludvigsson
JF
,
Andersson
E
,
Ekbom
A
, et al
.
External review and validation of the Swedish national inpatient register
.
BMC Public Health
2011
;
11
:
450
40.
Schmidt
M
,
Schmidt
SAJ
,
Sandegaard
JL
,
Ehrenstein
V
,
Pedersen
L
,
Sørensen
HT
.
The Danish National Patient Registry: a review of content, data quality, and research potential
.
Clin Epidemiol
2015
;
7
:
449
490
41.
Bakken
IJ
,
Ariansen
AMS
,
Knudsen
GP
,
Johansen
KI
,
Vollset
SE
.
The Norwegian Patient Registry and the Norwegian Registry for Primary Health Care: research potential of two nationwide health-care registries
.
Scand J Public Health
2020
;
48
:
49
55
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.